Sensitivity, Specificity of Vessel Involvement in Pancreatic Head Adenocarcinoma by Imaging Modality per Treatment Effect

Introduction

Pancreatic cancer is the fourth leading cause of cancer- related death in the United States and has among the poorest survival rates after diagnosis [1]. Even with successful resection, the five-year survival is only 15-25% [1].

Approximately 20% of patients have a resectable disease at the time of diagnosis and the remaining patients either have a metastatic or locally advanced disease at presentation [2, 3, 4]. Resectability in patients with newly diagnosed pancreatic ductal adenocarcinoma (PDAC) is determined by radiologic imaging demonstrating tumor proximity to or involvement of the mesenteric and portal vascular structures, in addition to exclusion of extra-pancreatic invasion of adjacent tissues and organs other than the duodenum [5, 6, 7, 8, 9, 10, 11, 12, 13, 14]. Vascular invasion is an important factor in determining margin status and resectability. The role of preoperative imaging is to select which patients are likely to have a margin free resection, and therefor likely to benefit from a pancreaticoduodenectomy [6, 8, 12]. Several classification schemes have been put forth by different organizations regarding vascular involvement and resectability of the peripancreatic vessels (celiac, hepatic, superior mesenteric artery, portal vein or superior mesenteric vein) including the National Compressive Cancer Network (NCCN), the Americas Hepato-Pancreato-Biliary Association (AHPBA) and the Society of Surgical Oncology (SSO) among others, which by and large, have overlapping criteria. Methods commonly used to assess the presence of vascular invasion by tumor include multi-detector computed tomography (MDCT), endoscopic ultrasonography (EUS), and magnetic resonance imaging (MRI) [7, 8]. Multi- detector computerized tomography has been regarded as the modality with the highest diagnostic accuracy for radiographic assessment of resectability. Zamboni et al demonstrated 100% sensitivity in the detection of resectability of pancreatic adenocarcinoma using MDCT, 94% specificity, 98% PPV, and 100% NPV [15]. A meta- analysis by Li et el demonstrated a diagnostic performance of CT with a sensitivity of 73% and a specificity of 95% in the evaluation of vascular invasion [6]. Furthermore, Lee et al demonstrated that MDCT has a sensitivity of 90%, specificity of 41%, PPV 85% and NPV 73% for overall tumor resectability [8]. Additionally, in their assessment of vascular involvement, a sensitivity and specificity of MDCT was calculated to be 61% and 96%, respectively [8]. MDCT of the pancreas is favorably complemented by EUS, which is more sensitive for the early detection of pancreatic lesions, and allows easy access to the pancreas for tissue diagnosis using fine needle aspiration (FNA), as well as assessing the relationship of the tumor to the critical vascular structures, with a reported sensitivity and specificity of 89% and 92% respectively for the major veins (SMV, PV, SPV), and 83% and 94% respectively for the major arteries (SMA, SPA) [7, 9, 11]. Current reports describe the sensitivity and specificity of these modalities on overall vascular involvement but do not specify the diagnostic accuracy of each modality on predicting involvement of the individual vascular structures that influence resectability. The purpose of our study was to describe the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), false positive rate (FPR), and false negative rate (FNR) of MDCT and EUS in predicting vascular involvement of the portal vein (PV), superior mesenteric artery (SMA) and superior mesenteric vein (SMV) in patients with PDAC in the setting of no prior neoadjuvant therapy and in the setting of prior neoadjuvant therapy.

Materials and Methods

Patients Data was collected retrospectively between August 2012 and September 2016. A total of 234 patient records were evaluated. We reviewed the surgical, pathologic, clinical and radiographic records of these patients from AdventHealth Orlando Hospital’s Cerner electronic medical record system.

Results

Of the 234 patient records evaluated, 85 (36%) patients met protocol criteria of diagnosis of PDAC who also underwent surgery. Of these 85 patients, 83 proceeded with pancreaticoduodenectomy and the remaining two were unrespectable as a result of vascular invasion (one patient had an unreconstructable SMV and another had involvement of the SMA). The mean age was 67 years. The male to female distribution was 49 (58%) to 36 (42%), respectively. The race distribution was as follows: 67 (79%) white, 6 (7%) black, 7(8%) Hispanic, 1(1%) Asian, and 4 (5%) other.

Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), false positive rate (FPR), and false negative rate (FNR) of each modality was calculated per vessel involved per treatment with the results noted in tables 1-5, whereby: ► Imaging Modality is CT, EUS, CT+EUS, intra-operative gross(IOG) evaluation by surgeon ► Treatment is

• without chemotherapy and without radiation
• with chemotherapy and without radiation
• with chemotherapy and with radiation ► Vessel involvement was quantified as degree of tumor involvement of vessel circumference. ► Vessels evaluated were:
• Superior Mesenteric Artery (SMA),
• Superior Mesenteric Vein (SMV),
• Portal Vein (PV)

Discussion

Accurate evaluation of vascular invasion in patients with pancreatic adenocarcinoma is very important in determining respectability of disease, which ultimately determines prognosis.

MDCT has been regarded as the modality with the highest diagnostic accuracy for radiographically assessing overall respectability with sensitivity as high as 100%, specificity of 94%, PPV of 98%, and NPV of 100% [15]. In the evaluation of vascular invasion, MDCT has a sensitivity of 73% and a specificity of 95% [6]. Furthermore, in the evaluation of vascular invasion, EUS has a sensitivity and specificity of 66% and 94% respectively [6]. EUS has a sensitivity and specificity of 89% and 92% respectively in the evaluation of major veins (SMA, PV, SPV), and 83% and 94% respectively in the evaluation of major arteries (SMA, SPA) [7].

Our study is unique in that it stratified the analysis by vessel type, imaging modality, and treatment setting. To our knowledge, there is no study to date that looks at sensitivity, specificity, PPV, NPV, FNR and FPR of CT and/or EUS with respect to vessel involvement (SMV, PV, SMA) in a given treatment setting (no treatment, neoadjuvant chemotherapy, neoadjuvant chemotherapy + radiation).

Our study suggests that in the setting of no treatment, CT has the highest sensitivity, specificity, PPV, NPV and the lowest FNR and FPR in the evaluation of SMA invasion by tumor: 22%, 94%, 50%, 83%, 16% and 50% respectively. Furthermore, in the setting of no treatment, EUS has the highest sensitivity, specificity, PPV, NPV and lowest FNR and FPR in the evaluation of SMV invasion by tumor: 33%, 95%, 33%, 95%, 4%, and 66% respectively. This data is in accordance with previous publications demonstrating higher sensitivity of CT for arterial evaluation and higher sensitivity of EUS for venous evaluation [6].

Our data shows that the modality with the highest specificity in the assessment of portal vein involvement is intra-operative evaluation regardless of treatment setting (95% for no treatment, 81% for chemotherapy only and 81% for chemotherapy + radiation). It is difficult to discern portal vein involvement radiographically given the close relationship of the portal vein with the normal pancreatic parenchyma without intervening fat along its right lateral and anterior margins. On anatomical grounds, it is therefore compelling to find that intra-operative evaluation is the ideal modality to ascertain portal vein invasion. One should therefore accept with caution the results of alternative modalities when it comes to the portal vein. Given an inadequate sample size, we cannot however comment on the ideal modality for evaluating PV involvement when it comes to sensitivity, PPV, NPV, FNR and FPR.

In the setting of neoadjuvant chemotherapy, or chemotherapy with radiation, our data suggests that either EUS or intra-operative evaluation is preferred over MDCT for the evaluation of SMA, SMV and PV involvement. This data is supported by previous studies demonstrating the limited role of CT in distinguishing between fibrosis and viable cancer post neoadjuvant therapy [5]. Diagnostic accuracy of CT for predicting resectability after neoadjuvant therapy has been documented in the range of 58% to 83% [13, 16]. It has been recognized that neoadjuvant therapy-induced tumor cell injury in pancreatic adenocarcinoma is mainly reflected by isovolumetric tissue replacement through fibrosis, rather than volume loss [10]. It is the lack of clear fat planes around critical vascular structures on post-neoadjuvant MDCT which can lead to an overestimation of unresectability.

The observed Pearson correlation between SMV and CA 19-9 may provide additional actionable information to the surgeon regarding resectability of disease where a higher level of CA 19-9 is moderately associated with greater degree of involvement of SMV by tumor. Furthermore, our study suggests a greater likelihood of SMV and PV involvement when SMA is involved by tumor. This information can be used by the surgeon in operative candidate selection by minimizing operative finding of unresectability [17].

In conclusion, our results imply that CT is the preferred modality for evaluation of SMA for tumor involvement, EUS is the preferred modality for evaluation of SMV, and the Portal Vein is best evaluated intra-operatively. Furthermore, EUS and intra-operative analysis are the preferred modalities over MDCT for re-evaluation of vessel involvement post- neoadjuvant therapy. We must use caution in applying our conclusions given the limited sample size. Our results do imply value in repeating the study with greater sample size.